Woman and a robot working in a car factory

We turn intelligence into motion. From robots to physical AI.

Your hand switches from a gentle grip on an egg to a firm hold on a box without a second thought. A robot needs tactile sensing, motor control and power working together to do the same. The AI model supplies the intelligence, and thousands of chips turn it into motion. We provide the electronics around the AI compute, from power, motor control, sensing and connectivity to safety and security.

Robotics
Article

Physical AI is AI that acts in the real world, through machines that perceive their surroundings, make decisions and move. Robotics is its most complex form.

Demographic change and labor shortages drive demand for autonomous robots. Developers train these robots for a wide range of use cases. They extend the capabilities of the workforce across sectors.

As demand and capability grow, annual production is expected to rise to millions of units within a decade.

1,650 GW waiting to connect
  • 100 million units by 2030 across all robot types, including widely adopted machines such as domestic and industrial robots (source: based on ABI Research 2025, Goldman Sachs 2025)
  • 6.5 million humanoid robots by 2035 (source: Goldman Sachs 2026)
  • Up to 300 million units by mid-century (source: UBS 2025)

Autonomy starts with the AI model. Intelligence lives in the model, the robot's brain, but the robot acts through the thousands of chips beneath it. These chips are the robot's nervous system, senses, muscles and immune system. They determine how precisely it moves, how long it runs on one charge and how safely it acts.

As robots take on more complex tasks, their architecture has to evolve with them. We understand the robot as one system and provide the electronics around the AI compute, from power, motor control, sensing and connectivity to safety and security. This represents a semiconductor value of up to 1,000 USD per robot for Infineon.

We help robots run longer, sense and move precisely, and act safely and securely, with trust built into the hardware.

Every robot is an integrated system of interdependent functional blocks, each with its own technical complexities, from dexterity and balanced movement to environmental perception, connectivity, functional safety, and security. Advanced semiconductors are needed to overcome every one of these challenges. Infineon helps customers navigate that complexity, from architecture development in the lab to bringing robots into real-world deployment.

Bidirectional Energy Flow

Challenge
Operating time decides how much work a robot gets done. A bigger battery stores more energy but adds weight. In flexible machines that move between workstations, low weight is a design precondition.

Solution
We solve this with efficiency along the entire power path. As the global number one in power semiconductors, we power the robot from the battery to the AI compute and every actuator. Our chips manage charging and discharging and regulate current, so less energy is lost and the robot runs longer on one charge. We match the right semiconductor material, Silicon, Silicon Carbide or Gallium Nitride, to each voltage and design requirement, and support the move to higher voltages, which deliver the same power at lower current and reduced energy loss. There is no AI without power. The power-hungry AI compute in a robot faces the same challenge as in an AI data center, delivering high power efficiently in a compact space. Drawing on our leading position in power for AI data centers, from the grid to the processor core, we bring this competency into robotics.

Bidirectional Energy Flow

Challenge
Robots work among people in factories, hospitals and homes. Trust has to be built into the hardware. It is the precondition for scale. Robots are also networked, software-defined systems. A cyberattack on a physical AI system can cause physical harm as well as data loss. Functional safety, cybersecurity and connectivity must be addressed together at chip level.

Solution
As the global number one in microcontrollers, we bring decades of industrial and automotive safety expertise to robotics, built to the highest standards. Our microcontrollers monitor the robot independently of the AI at the edge and help bring it into a safe state. Our post-quantum-ready security solutions comply with regulatory frameworks and enable secured authentication, updates and communication. Our security net covers every data path, while our connectivity solutions carry the data through the robot. We offer Ethernet today, a proven standard in automotive with high bandwidth and low latency, now increasingly entering robotics. We also connect the robot to other devices and the cloud with Wi-Fi and Bluetooth for smooth machine interaction.

Private Green energy generation

Challenge
What a robot perceives and feels ultimately has to become movement. This happens in the actuators, the robot’s joints, that turn electrical energy into motion. A humanoid, the most complex form in robotics, has more than 70 actuators, and each needs precise control in a compact form. 

Solution
We lead in motor control for robot actuators of every size. Our integrated system solutions combine sensing, microcontrollers for zonal compute and control, and power in one solution. The power semiconductors are tailored to the actuator type, with Gallium Nitride specifically suited to compact actuators and dexterous hands, and Silicon and Silicon Carbide to heavy-duty actuators that lift. This complete reference design is integration-ready for robot developers, reducing design effort and enabling precise, stable and dexterous movement.

Data Center

Challenge
Every action starts with awareness, and awareness is built on sensors. The dexterous hand sets the hardest sensing task. It must feel pressure, texture and slip, and it must adapt from a gentle touch to a firm hold.

Solution
Radar, cameras and microphones let a robot see and hear, also in low light. Our environmental sensing solutions have a proven track record across automotive, industrial and consumer applications. In tactile sensing, we are ahead. Our best-in-class solutions enable the development of hands with human-like grasping. Every sensor delivers different data. Sensor fusion merges all of it into one reliable picture. The central compute unit then uses this picture for decision.

Robot architectures are evolving fast, and we shape them with customers and partners. Our portfolio depth and system understanding let us co-engineer from prototype to mass production.

  • Our solutions are integrated into digital twin environments, so developers can test virtually before a single robot is built.
  • Our diversified manufacturing and a resilient global supply chain support volume ramp-up.
  • A strong partner network brings solutions to market at scale.
  • In our own fabs, we explore physical AI and capture requirements for semiconductor manufacturing with robot developers.

Chip makers, AI players, robot makers, startups and researchers form one ecosystem. Together we set common standards, drive innovation and make physical AI real at scale, with trust built in.

Which robot types does this technology support, and will humanoids dominate?

The task defines the robot, and there is no universal robot. Cobots and wheeled robots are often preferred in industry for stability and efficiency, while humanoids offer flexibility in human-centric environments. A mix is usually most effective. Our electronics foundation applies to industrial robots, cobots, mobile and domestic robots, drones, eVTOLs and humanoids. The humanoid is the most complex case, and the same architecture scales to every other robot type.

Where is a robot's AI trained and why is there no AI without power?

Robots are trained in AI data centers, the training ground of physical AI, like a robot academy where they prepare for diverse use cases. In operation, the intelligence runs inside the robot, at the edge, under battery, thermal and load constraints. Runtime therefore comes down to efficiency. We power AI on both levels, from the grid to the processor core in the data center, and from the battery to the central AI compute and every actuator in the robot.

How is safety organized inside a humanoid robot?

Safety works on two levels. At system level, a dedicated microcontroller acts as safety companion to the central compute unit. It validates safety-critical decisions and keeps the deterministic path to a safe, stable condition, which goes beyond a safe stop. At actuator level, distributed microcontrollers process data in real time. Functional safety covers faults, while cybersecurity covers intentional interference. Both form the foundation of trust.

How does a robot make sure it only runs trusted software?

Security is rooted in hardware. A trusted platform module gives the robot a secured built-in identity and enables secured boot, so only authorized software executes. Authentication and encryption protect updates and communication. Our certified modules provide security across the product lifecycle. The same trust covers every path data takes, from the wired in-robot network to the wireless links to the cloud.

What happens inside a single robot joint?

Each joint needs precise, efficient motor control. In the actuator, power semiconductors, microcontrollers, position and current sensors work together, controlling joint movement and regulating the electrical flow driving the motor. Together they enable torque sensing for whole body control, which recognizes internal and external force for smooth motion and balancing. This is the foundation of balanced movement and action in robotics. 

How our deep portfolio and system expertise benefit you